Non-Separable Primary Transforms for High-Resolution Image Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies face challenges in efficiently encoding and decoding high-resolution and high-quality images, particularly in handling transform processes with separable primary transforms that do not fully optimize encoding efficiency.
Innovation Solution
The use of a non-separable primary transform kernel, determined based on encoding parameters, for both encoding and decoding processes, which allows for improved transform performance and encoding efficiency by applying conditions specific to luma and chroma components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If separable primary transform is used for image compression, then device complexity is reduced and ease of manufacture is improved, but encoding efficiency deteriorates for high-resolution images
Solution Approach 1:
The transform process is segmented into separable horizontal and vertical transforms. Instead of applying a single complex non-separable transform, the patent divides the transformation into two independent one-dimensional transforms that can be applied sequentially, reducing computational complexity while maintaining compression effectiveness.
Solution Approach 2:
The patent transitions from a two-dimensional non-separable transform to two separate one-dimensional transforms applied in different dimensions (horizontal and vertical). This dimensional decomposition simplifies the transform kernel requirements and reduces computational burden while preserving the ability to compress high-resolution images effectively.
2Productivity
If non-separable primary transform kernel is used, then transform performance and encoding efficiency are improved, but device complexity increases
Solution Approach 1:
The complex non-separable transform kernel is segmented into simpler separable transform kernels for horizontal and vertical directions. This segmentation reduces the mathematical complexity of the transform while maintaining the ability to effectively compress image data by capturing correlations in both dimensions.
Solution Approach 2:
The patent changes the parameters of the transform by selecting different transform kernels (e.g., DCT, DST) for horizontal and vertical directions based on prediction modes and block characteristics. This parameter adaptation allows the system to achieve high encoding efficiency without requiring a single complex non-separable kernel, thus reducing overall device complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image decoding method and device according to the present disclosure may: acquire residual information from a bitstream; derive transform coefficients of a current block on the basis of the residual information; perform at least one of inverse quantization or inverse transform on the transform coefficients of the current block and derive residual samples of the current block; and reconstruct the current block on the basis of the residual samples of the current block. Here, the inverse transform may be performed on the basis of an inverse non-separable primary transform.